The Height of Dutch Children from 1955 to 2009

nature publishing group Population Study Articles

The world’s tallest nation has stopped growing taller: the height of Dutch children from 1955 to 2009

Yvonne Schönbeck1, Henk Talma2, Paula van Dommelen3, Boudewijn Bakker4, Simone E. Buitendijk1, Remy A. HiraSing2 and Stef van Buuren3

Background: Records show that mean height in The but studies of trends in mean height and of height variations in Netherlands has increased since 1858. This study looks at tall populations could suggest a possible direction. whether this trend in the world’s tallest nation is continuing. The Dutch population is the tallest in the world and has been We consider the influence of the geographical region, and of well studied. Before 1955, population height in The Netherlands the child and parental education, on changes in height. was estimated on the basis of data for conscripts, i.e., men, and Methods: We compared the height of young Dutch people nonrandom samples from the population. Between 1955 and aged 0–21 y as determined on the basis of the growth study of 1997, there were four large cross-sectional nationwide growth 2009, with the height data from growth studies conducted in studies. These studies showed that young Dutch adults are among 1955, 1965, 1980, and 1997. the tallest people in the world, with women measuring almost Results: The analysis sample included 5,811 boys and 6,194 171 cm on average and men 184 cm on average in 1997 (4). girls. Height by age was the same as in 1997. Mean final height In The Netherlands, people have been getting taller since was 183.8 cm (SD = 7.1 cm) in boys and 170.7 cm (SD = 6.3 cm) 1858 (2,5). At that time, conscripts were an average of 163 cm. in girls. The educational levels of both children and their par- This indicates a height gain of 21 cm over a period of 140 y. In ents are positively correlated with mean height. Since 1997, The Netherlands, as in most Northern European countries, the differences between geographical regions have decreased upward trend is slowing down significantly (2,6), and it even but not vanished, with the northern population being the appears to be reversing in the United States (7). tallest. This article presents new data about height from the Fifth Conclusion: The world’s tallest population has stopped Dutch Growth Study, which was carried out in 2009. The main growing taller after a period of 150 y, the cause of which is research questions are: Is the 140-y trend of increasing height in unclear. The Dutch may have reached the optimal height distri- the world’s tallest nation still continuing? If not, has there been bution. Alternatively, growth-promoting environmental factors a compression of height variation? Do geographical region and may have stabilized in the past decade, preventing the popula- educational levels affect the development of mean height? tion from attaining its full growth potential. Results ver the past decades, human height has increased in most The sample comprised of 10,030 children of Dutch origin and Oindustrialized countries. This trend has been the result 1,975 imputed cases, resulting in an analysis sample of 12,005 of improvements in the nutritional, hygienic, economic, and children (5,811 boys and 6,194 girls) aged 0–21 y. Table 1 lists health status of the populations in question (1). Such factors the numbers of children in each age group. may allow disadvantaged individuals in particular to grow Height by age and sex was normally distributed, and there- taller. However, we have not found any reduction in height fore standard deviation scores (SDS) could easily be calculated variation (2). It appears that individuals benefit more or less as (height − mean height)/SD. Table 2 summarizes the mean equally from improvements in living conditions. Do we believe and SD for height extracted from the height references, clas- that we could all be giants if living conditions for humans were sified according to age and sex. Extended tables are available to improve further, or is there some maximum (and presumably from the authors on request. optimal) population average beyond which our species will not New growth diagrams were constructed for boys and girls grow, no matter how favorable the circumstances? Although aged 0–21 y, including 2.5, 2, 1, 0, −1, −2, −2.5, and −3 SD being tall is often associated with being healthy, this may not lines. The −3 SD line was added for practical reasons because always be the case. Taller adults have, for example, higher rates the Dutch referral criteria require children with a height SDS of several types of cancer (3). The question of whether there is of less than −3 SD to be referred for short stature. Figure 1a,b a maximum population average is almost impossible to answer, shows the new growth diagrams for boys and girls aged 1–21 y.

1Department of Child Health, Netherlands Organisation for Applied Scientific Research (TNO), Leiden, The Netherlands; 2Institute for Health and Care Research (EMGO+), Free University Medical Center (VUmc), Amsterdam, The Netherlands; 3Department of Life Style, Netherlands Organisation for Applied Scientific Research (TNO), Leiden, The Netherlands; 4Department of Pediatrics, Leiden University Medical Center, Leiden, The Netherlands. Correspondence: Yvonne Schönbeck ([email protected]) Received 5 December 2011; accepted 18 October 2012; advance online publication 30 January 2013. doi:10.1038/pr.2012.189

Table 1. Number of observations per age and sex in the Fifth Dutch Table 2. Mean height and SD extracted from height references for Growth Study 2009 boys and girls per age Age (y) Boys (n) Girls (n) Height 0–1 999 992 Boys Girls 1–2 536 543 Age (y) Mean SD Mean SD 2–3 343 349 0.0767a 54.4 2.3 53.7 2.3 3–4 311 308 0.2500 60.9 2.4 59.7 2.4 4–5 138 184 0.5000 68.0 2.5 66.4 2.5 5–6 187 214 0.7500 72.9 2.6 72.9 2.6 6–7 161 210 1.0000 76.7 2.7 75.0 2.8 7–8 202 224 1.2500 79.9 2.8 78.4 2.9 8–9 226 246 1.5000 82.8 2.9 81.5 3.1 9–10 233 254 1.7500 85.6 3.1 84.4 3.2 10–11 318 367 2.0000 88.4 3.2 87.1 3.3 11–12 280 287 3.0000 97.8 3.7 97.0 3.8 12–13 255 294 4.0000 105.5 4.2 104.9 4.2 13–14 196 235 5.0000 113.2 4.7 112.1 4.7 14–15 256 307 6.0000 119.9 5.1 118.8 5.1 15–16 245 265 7.0000 126.2 5.4 125.3 5.4 16–17 163 178 8.0000 132.5 5.8 131.3 5.8 17–18 205 179 9.0000 138.5 6.2 137.3 6.1 18–19 180 185 10.0000 143.7 6.6 143.5 6.5 19–20 166 158 11.0000 149.0 7.0 149.7 6.7 20–21 137 165 12.0000 155.2 7.6 155.7 6.9 21–22 74 50 13.0000 161.8 8.2 160.8 6.9 Total 5,811 6,194 14.0000 168.5 8.3 164.5 6.9 15.0000 175.2 7.9 166.9 6.8 Similar growth diagrams were established for children aged 16.0000 179.1 7.6 168.3 6.7 0–15 mo and 0–4 y. Note that there is no visual evidence of the 17.0000 181.0 7.4 169.2 6.6 pubertal height growth spurt in the growth diagram for girls. 18.0000 182.4 7.2 169.7 6.5 Figure 2 shows the trend in the mean final height in boys and 19.0000 183.6 7.1 170.1 6.4 girls from 1955 to 2009. The figure illustrates that the upward secular trend in the mean final height of young Dutch adults 20.0000 183.8 7.1 170.5 6.3 has stopped. Mean final height was 183.8 cm (SD = 7.1 cm) in 21.0000 183.8 7.1 170.7 6.3 Dutch boys and 170.7 cm (SD = 6.3 cm) in Dutch girls. These a0.0767 = 4 wk. heights do not differ significantly from the heights in 1997 (184.0 cm for boys and 170.6 cm for girls). Figure 2 indicates Figure 4 presents the mean height SDS of Dutch children that the pace of the increase in secular height over the years aged 0–21 y by geographical region, level of child education, slowed down both in boys and girls, falling to almost zero in and level of parental education in 1997 and 2009, corrected for 2009. each of the two other factors. Variation across the geographi- Figure 3 shows, for Dutch children aged 1–21 y, the height cal regions has lessened since 1997. The mean height SDS difference with respect to 1955 in 1965, 1980, 1997, and 2009 in the north and in the south are closer to the overall mean for each age. The figure clearly shows an increase in height for (Figure 4a). The mean height SDS in the north of the country the period 1955–1997, especially from the age of 4 y onward, is still significantly higher than that in the south (difference = with the increase being more marked in boys than in girls. This 0.22 SDS; P < 0.001) and the major cities (0.12 SDS; P = 0.018). figure also shows that the secular trend has stopped since 1997, This means that 21-y-old boys and girls from the north are on not only for final height but also for all ages, with the 1997 and average 1.6 and 1.4 cm taller, respectively, than those from the 2009 lines being a close match for all ages. There is only a small south. Figure 4b,c shows that, in 2009, as in 1997, mean height difference in height at the ages of 11–13 y in boys and girls SDS increased in line with the educational levels of children and and at the age of 15 y in boys. These differences are more pro- their parents. Since 1997, mean height SDS has not changed nounced in boys (0.7–1.2 cm) than in girls (0.4–0.5 cm). This significantly. Nevertheless, the mean height SDS of children may indicate slightly earlier puberty. with lower levels of education was significantly lower than that

Figure 1. New Dutch growth diagrams, including height by age references for (a) boys and (b) girls aged 1–21 y.

185 educated parents are on average 0.8 cm shorter than those of higher-educated parents. For girls, this difference is 0.7 cm. 180 Discussion 175 This study shows that the height of Dutch children in 2009 is very similar to that in 1997. There was a slight increase in mean 170 height in early puberty in boys and girls, indicating that they

Final height (cm) were taller at a younger age than in 1997. However, this did 165 not affect the final height, as no significant differences in final height were seen as compared with 1997. There was no com- 160 pression of height variation observed. 1955 1965 1980 1997 2009 It is remarkable that the secular height trend spanning 150 Study year y has come to an end, at least temporarily. The cause is not yet Figure 2. Mean final height in Dutch boys (filled circles) and girls (open clear. Economic factors are not likely to be involved because circles) in 1955–2009. gross domestic product in The Netherlands did not level off; indeed, it rose exponentially between 1920 and 2008 (8). We of those with higher levels of education in 2009 (P = 0.003), also looked at the course of the Life Situation Index (LSI) whereas this difference was nonsignificant in 1997 (P = 0.438), between 1974 and 2008. This index was established by The indicating that 21 y olds with lower levels of education are on Netherlands Institute for Social Research to measure the prog- average 1.4 cm (boys) and 1.2 cm (girls) shorter than those with ress of Dutch society using indicators that go beyond mere higher levels of education. At the age of 21 y, boys of lower- economic growth. The index incorporates indicators in eight

a 12

6 erence from 1955 (cm) 4

Height diff 2

0 132 45678910 11 12 13 14 15 16 17 18 19 20 21 Age (years)

b 10

4 erence from 1955 (cm) 2

Height diff 0 2345678910 11 12 13 14 15 16 17 18 19 20 21 Age (years) –2

Figure 3. Height increase since 1955 by age in Dutch (a) boys and (b) girls in 2009 (solid line), 1997 (striped line), 1980 (dotted line), and 1965 (gray striped lines). domains: health, sport, social participation (loneliness, vol- diagnosed growth disorder or who were on growth-interfering unteering), cultural/leisure activities, housing, mobility, holi- medication were excluded from the sample. It is possible that days, and possession of assets (9,10). The LSI rose substantially some growth disorders and growth-interfering medication between 1974 and 2008. Between 1990 and 2008, the differ- were not mentioned by a child or parent, or were not in the ences in LSI scores between socioeconomic classes decreased. medical records and therefore were missed by the health-care This was mainly due to a higher rise in the LSI score for lower provider. This could lead to an underestimation of mean height socioeconomic groups (9). A correlation between height per age. However, we believe that even if we missed some chil- and LSI has not been demonstrated, but one would expect dren, they would form only a very small part of the sample a positive correlation, as with economic growth. The rise in size per age. Moreover, the same method for inquiring about LSI mentioned here and the reduction in the variation in LSI and exclusion of growth disorders and medication was applied scores across socioeconomic groups do not therefore explain in 1997. Therefore, it is unlikely that growth disorders and the break in the Dutch height trend. Children of parents born growth-interfering medication influenced the height differ- outside The Netherlands were excluded from the analyses. The ence between 1997 and 2009. so-called third-generation immigrants—those with parents There will always be some differences in, for example, the born in The Netherlands but grandparents born outside The genetic background and socioeconomic status in a given Netherlands—were not excluded. This group makes up an population, leading to height variations. Given that the Dutch increasing proportion of the population in The Netherlands are the tallest people in the world, and that they are no longer but is still relatively small. On the basis of data on the country growing taller, we could hypothesize that the mean height of of birth of the grandparents, only 4–5% of our study population the population has reached the maximum possible. It could were from the third generation. Although the third-generation be that the distribution of population characteristics in The immigrants may be partly accountable for the flattening of the Netherlands, including environmental factors, is the “best” Dutch growth trend, this group is too small to fully explain achievable at the country level, leading to a mean population the stagnation in height. Moreover, children known to have a height of 184 cm for boys and 171 cm for girls.

a 0.3 for monitoring child height over time, but one should bear in

0.2 mind that these charts do not fully represent how an individual child typically grows. 0.1 Only children with both parents born in The Netherlands 0.0 were included. The selection of children of Dutch decent may –0.1 complicate comparisons of height/final height between coun- Height SDS –0.2 tries because some countries include all ethnic groups in their data. On the basis of data for other ethnic groups in the dataset –0.3 North East West Major South cities of the Fifth Dutch Growth Study, we estimated an adjusted final height that includes other ethnic groups. The adjusted mean b 0.4 final height would be 183.8 − 0.4 = 183.4 cm in boys and 170.7 0.3 − 0.3 = 170.4 cm in girls. These numbers are an estimate of final 0.2 height for all children living in The Netherlands, irrespective of 0.1 ethnicity. As the differences between the Dutch and the non- 0 Dutch groups are substantial, separate growth references were Height SDS –0.1 constructed for children of Moroccan and Turkish decent, the two largest minorities in The Netherlands. These will be - pre –0.2 Low Middle High sented elsewhere. We found that the height difference between the north and c 0.4 the south of The Netherlands has diminished since 1997. This 0.3 was seen irrespective of child and parental educational levels. 0.2 This change could be explained by factors such as fewer health 0.1 inequalities or changes in national motility, in other words 0 more migration between geographical regions. However, we Height SDS found no evidence indicating a reduction of health inequali- –0.1 ties between these regions, and there have been no major shifts –0.2 Low Middle High in motility in The Netherlands since 1997 that could explain Figure 4. Differences in height SDS by (a) geographical region, (b) edu- the phenomenon (17). Random sampling variation between cational level of the child, and (c) highest completed education of parents the 1997 and 2009 study may have contributed to the smaller between 1997 (open circles) and 2009 (filled circles). Values are means, north–south difference. However, both studies were nationally adjusted for the effects of the other factors. SDS, standard deviation scores. representative. The study population of the Fifth Dutch Growth Study was large and randomly selected, and advanced imputa- Alternatively, growth-promoting environmental factors may tion techniques were carried out for underrepresented groups have stabilized in the past decade, preventing the Dutch from to obtain a representative sample. Although we do not have a fulfilling their full growth potential. We could hypothesize that firm explanation for the reduction in height inequality between current lifestyles have a negative effect on height. Komlos (11), the north and the south of The Netherlands, the 1.6 cm (boys) who looked at differences in height and BMI between Dutch and 1.4 cm (girls) difference in 21 year olds seem very plausible and US children, also suggested this. Where insufficient food because they are similar to the 2 cm north–south difference intake may have led to stunting decades ago, easy access to found in a large Dutch study based on self-reported growth food/fast food nowadays leads to unhealthy eating habits. For data (18). example, the high consumption of milk in The Netherlands, It is remarkable that the difference in mean height SDS for which has been linked to tallness (12), declined over the past children with lower and higher education levels has increased decade from 63 l per capita per year in 2000 to 60 in 2010 (13). since 1997. This suggests that the inequality associated with Unhealthy eating habits may lead to inadequate nutrient intake, education has increased. We found no explanation for this which may result in lower height. Furthermore, an unhealthy in the literature of national registries. It should be noted that diet in combination with less energy expenditure due to a sed- there have been no significant changes in the mean height SDS entary lifestyle leads to an increase in overweight and obesity, in any of the categories since 1997. Monitoring this trend is a phenomenon that has also been observed in Dutch children important because larger height differences can be a sign of (14). Higher BMI is associated with earlier onset of matura- increasing health inequalities. tion and menarche, which, in turn, are related to lower height The upward secular height trend has slowed in most Northern (15,16). If our hypothesis is valid, adopting healthier lifestyles European countries (2,6). Conscript data show that height in (including healthier diet and more activity) could result in the Danish and Norwegian men remained stable between 2000 reemergence of the positive height trend in the future. and 2009, and an earlier study based on self-reported data from No pubertal height growth spurt can be observed in the growth Dutch adults also indicated that the secular increase in height diagram for girls. This is due to the fact that Dutch growth studies had come to a halt (18–20). Despite the stagnating secular are cross-sectional studies. These growth diagrams are suitable trend, the Dutch are still the tallest people in the world, followed

Copyright © 2013 International Pediatric Research Foundation, Inc. Volume 73 | Number 3 | March 2013 Pediatric Research 375 Articles Schönbeck et al. directly by the Danish, who are reported to have a mean height characteristics were obtained from the parents or from adolescents of 183.7 cm (men) and 168.6 cm (women), and the Norwegians, themselves by health-care professionals using a questionnaire. with averages of 182 cm (men) and 169 cm (women) (21,22). Variable Definitions Pineau et al. (23) found boys in the Dinaric Alps to be 1.5 cm The sample was divided into five geographical regions: North (Friesland, taller than their Dutch counterparts, whereas girls reach heights Groningen, Drenthe), East (Overijssel, Gelderland, Flevoland), West similar to their Dutch counterparts. Although this study looks (Noord-Holland, Zuid-Holland, Utrecht – not including the four major cities), South (Zeeland, Noord-Brabant, Limburg), and the four major at a region rather than a country, and corrects boys’ height for Dutch cities (Amsterdam, Rotterdam, Utrecht, and The Hague). The growth after the age of 17 y, it indicates that young adults from educational level of the child was determined at the time of measure- this region in Eastern Europe are also tall. In fact, men in the ment. If an adolescent older than 15 y had left the educational system, the highest completed education was recorded. The educational level of Balkans were already among the tallest in Europe in the 19th the parents was defined as the educational level of the highest educated century (11). Future studies of population height are expected parent and broken down into low, middle, high, and unknown (29). For to reveal whether the maximum mean population height is the purposes of data cleaning and the comparison of height changes indeed 184 cm for boys and 171 for girls. The new Dutch refer- between 1997 and 2009, SDS were calculated for each age using the 1997 height references (2,30). ence charts are available at http://www.tno.nl/growth. Statistical Analyses Methods Data were cleaned using descriptive statistics that included fre- Data Sources quency tables, contingency tables, and scatter plots. Outliers, defined Data for the earlier Dutch growth studies were collected in 1955, as values more than +5 SDS or less than −5 SDS, were checked for 1965, 1980, and 1997 and published elsewhere (2,24–26). The Fifth data entry errors and corrected wherever possible. If no correction Dutch Growth Study was a cross-sectional study that collected data was possible, these measurements were considered erroneous and for a representative sample of children aged 0–21 y in The Netherlands defined as missing. The distributions for child and parental educa- between May 2008 and October 2009. The sample was stratified by tional levels and for geographical regions were compared with the region (regions of Municipal Health Services), sex, and age in line with national distributions (31) to check for representativeness. We used national distributions (27). Until the age of 4 y, measurements were per- the multivariate imputation by chained equation method (32,33) formed during regular periodical health examinations in 28 Well Baby to correct for educational levels or geographical regions that were Clinics at the ages of 1, 2, 3, 6, 9, 12, 15, 18, 21, 24, 30, 36, and 45 mo. underrepresented. The imputation model was based on age, sex, Between the ages of 4 and 8 y, children were measured at 23 Municipal height SDS, weight SDS, waist SDS, hip SDS, parental height, birth Health Service offices during two regular preventive health assessments weight, ethnicity, socioeconomic status score (34), educational levels performed at the ages of ~5.5 and 7.5 y. From the age of 9 y onward, of the children and parents, and geographical region. Passive imputa- children were invited personally by the Municipal Health Services after tion was performed on the interactions between age and region, age being selected in an age- and sex-stratified random sampling procedure and all SDS, age squared, and age squared and all SDS. There was no from the registers of the Municipal Register Office. Furthermore, we adjustment for the transition from length to height measurements, conducted measurements at randomly selected primary and second- given that the age of the transition from length to height measure- ary schools, at two colleges and two universities across the country, ment differs from child to child in daily practice. Length and height and at a national youth festival. The collection was supplemented by were smoothly joined. data from two recent large, high-quality studies performed by trained SDS reference values for height by age in 2009 were calculated using staff at primary schools in Amsterdam (Municipal Health Services in the LMS method (35). For these calculations, the data of 146 children Amsterdam) and at a vocational education institution in the east of aged 22–25 y were added to the dataset to obtain stable-fit results at The Netherlands (Deltion College Zwolle/OPOZ VU-Windesheim). the age of 21 y. These subjects were excluded from further analysis. The Random samples from these data (n = 270 and n = 342, respectively) LMS method calculates three smooth curves representing skewness (L were added to the data for this study. curve), the median (M curve), and coefficient of variation (S curve) to establish SDS lines. L values of 1 indicate normality, and smaller values Ethics Statement represent progressively greater skewness. The M curve is the 0 SDS Data collection for growth studies is a part of routine youth health care line or 50th centile curve. The S curve defines the coefficient of varia- in The Netherlands and is not regarded as medical research. Before the tion, and when multiplied by 100, it can be interpreted as a percentage. measurement, oral consent was obtained from each child (and from The smoothing parameters (effective degrees of freedom) for the L, the parent when children were younger than 16 y). Cooperation, or M, and S curves were chosen by creating worm plots: local detrended lack thereof, was recorded on the questionnaire. Data were analyzed quantile-quantile plots of the SDS of the reference samples across 16 anonymously. The Medical Ethics Review Board of Leiden University age groups (36). A log transformation was applied to the age axis. The Medical Center approved the study and the manner in which consent curves were fitted as cubic splines. Finally, estimates were made of the was obtained. age-related SDS reference values for height in 2009. Differences in mean heights between the samples of 2009, 1997, Exclusion Criteria 1980, 1965, and 1955 were computed for each age group. “Mean Exclusion criteria for the Fifth Dutch Growth Study were similar to height” in this article refers to the 50th percentile of the reference those for the previous Dutch growth studies: children with diagnosed charts, unless otherwise indicated. Children from the 2009 study with growth disorders and those on growth-interfering medication were a birth weight of <2,500 g were not excluded from the comparisons excluded (2.4% of the total population). Children with one or two par- because their inclusion has only a slight effect on mean height by age ents born outside The Netherlands (28) were excluded from the analy- and because, as in 1997, they were included in the calculation of the ses presented in this article. Results with respect to children of Turkish height references. and Moroccan origin will be described elsewhere. Multivariate analysis (ANOVA) was conducted to assess the associ- ations and trends for the demographic variables (geographical region, Measurements child educational level, and parental educational level) with height Trained health-care professionals carried out the standardized mea- SDS in 2009 and differences between 1997 and 2009. Consequently, surements. Infant length was measured to the nearest 0.1 cm in the the values were corrected for the effects of the other two factors. supine position until 2 y of age. From 2 y of age onward, standing R version 2.9.0 (R development core team, Vienna, Austria) with height was measured to the nearest 0.1 cm. The children’s demographic the GAMLSS package (37) was used for imputation and for estimating

376 Pediatric Research Volume 73 | Number 3 | March 2013 Copyright © 2013 International Pediatric Research Foundation, Inc. Height of Dutch children from 1955 to 2009 Articles the height SDS reference values. All other statistical analyses were 16. Freedman DS, Khan LK, Serdula MK, Dietz WH, Srinivasan SR, Berenson performed in SPSS version 17.0 for Windows (SPSS, Chicago, IL). GS. The relation of menarcheal age to obesity in childhood and adulthood: the Bogalusa heart study. BMC Pediatr 2003;3:3. ACKNOWLEDGMENTS 17. Centraal Bureau voor de Statistiek, (Statistics Netherlands). [Population We thank all the children, their parents, and the community health Development CBS]. (http://statline.cbs.nl/StatWeb/selection/?DM=SLNL care workers who participated in this study. We thank ActiZ and &PA=37259NED&VW=T). Accessed 7 July 2011. GGD-Nederland for their cooperation in the study. We thank Deltion 18. Frenken F. [Deceleration of Height and Weight Increase]. Voorburg/Heer- College/OPOZ VU-Windesheim and GGD Amsterdam for kindly provid- len, The Netherlands: Centraal Bureau voor de Statistiek (Statistics Nether- ing their data sets. lands), Demographic Statistics, 4th quarter 2007, 2007. Statement of Financial Support 19. Denmark Statistics. Population and elections. In: Agerskov U, Bisgaard MP, This study was funded by the Dutch Ministry of Welfare and Sport (grants eds. Statistical Yearbook 2011. Copenhagen, Denmark: Statistics Denmark, 310434, 312617, and 315319). The researchers were independent of the 2011:53. funder. 20. Norway Statistics. Statistical Yearbook of Norway 2010 - Health and Social Conditions. (http://www.ssb.no/english/yearbook/2010/tab/tab-108.html). REFERENCES Accessed 3 May 2012. 1. Hauspie RC, Vercauteren M, Susanne C. Secular changes in growth. Horm 21. Garcia J, Quintana-Domeque C. The evolution of adult height in Europe: a Res 1996;45:Suppl 2:8–17. brief note. Econ Hum Biol 2007;5:340–9. 2. Fredriks AM, van Buuren S, Burgmeijer RJ, et al. Continuing positive secular 22. Norway Helsedirektoratet. [Physical Fitness Among Adults and the Elderly growth change in The Netherlands 1955-1997. Pediatr Res 2000;47:316–23. in Norway - Results From a Survey in 2009-2010]. Oslo, Norway: Helse- 3. Batty GD, Shipley MJ, Gunnell D, et al. Height, wealth, and health: an direktoratet Norway, 2010. overview with new data from three longitudinal studies. Econ Hum Biol 23. Pineau JC, Delamarche P, Bozinovic S. [Average height of adolescents in 2009;7:137–52. the Dinaric Alps]. C R Biol 2005;328:841–6. 4 Wieringen van JC. Secular growth changes. In: Falkner F, Tanner JM, eds. 24. Wijn de JF, Haas de JH. [Growth Diagrams for Ages 1-25 Years in The Neth- Human Growth. New York: Plenum Press, 1988:307–31. erlands]. Leiden, The Netherlands: Nederlands Instituut voor Praeventieve 5. Drukker J, Tassenaar PG. Paradoxes of modernization and material well- Geneeskunde, 1960. being in The Netherlands during the 19th century. In: Steckel RH, Floud 25. Wieringen van JC, Wafelbakker F, Verbrugge HP, Haas de JH. [Growth R, eds. Health and Welfare During Industrialization. Chicago: Chicago Diagrams 1965 The Netherlands]. Leiden/Groningen, The Netherlands: University Press, 1997:331–79. Nederlands Instituut voor Praeventieve Geneeskunde/Wolters Noord- 6. Gohlke B, Woelfle J. Growth and puberty in German children: is there still hoff, 1965. a positive secular trend? Dtsch Arztebl Int 2009;106:377–82. 26. Roede MJ, Wieringen van JC. [Growth diagrams 1980: Netherlands third 7. Komlos J. The recent decline in the height of African-American women. nationwide survey]. Tijdschr Soc Gezondheidsz 1985;63:1–34. Econ Hum Biol 2010;8:58–66. 27. Centraal Bureau voor de Statistiek, (Statistics Netherlands). [Demographic 8. Centraal Bureau voor de Statistiek, (Statistics Netherlands). [Dutch Gross statistics 2006]. (http://statline.cbs.nl/StatWeb/selection/?DM=SLNL&PA Domestic Product 2000-2009]. (http://statline.cbs.nl/StatWeb/publication/? =37713&VW=T). Accessed 12 November 2007. VW=T&DM=SLNL&PA=71015NED&D1=0&D2=0&D3=a&HD= 28. Keij I. [Standard definition immigrants]. Index 2000;10:24–5. 120503-1943&HDR=T&STB=G1,G2). Accessed 3 May 2012. 29. Verweij A. [Categorising educational level]. In: Volksgezondheid Toekomst 9. Boelhouwer J. [Life situation, happiness and quality of life]. In: Bijl R, Verkenning, ed. [The Dutch 2008 Public Health Status and Forecast Report]. Boelhouwer J, Cloïn M, Pommer E, eds. [The Social State of The Nether- Bilthoven, The Netherlands: RIVM, 2008. lands 2011]. Den Haag: Sociaal Cultureel Planbureau, 2011:307. 30. Cole TJ. The LMS method for constructing normalized growth standards. 10. Boelhouwer J. [Life situation]. In: Roes T, ed. [Social State of The Nether- Eur J Clin Nutr 1990;44:45–60. lands 2001]. Den Haag: Sociaal Cultureel Planbureau, 2001:241. 31. Centraal Bureau voor de Statistiek, (Statistics Netherlands). [Demographic 11. Komlos J. Anthropometric evidence on economic growth, biological well- statistics 2008]. (http://statline.cbs.nl/StatWeb/selection/?DM=SLNL&PA= being and regional convergence in the Habsburg Monarchy, c. 1850-1910. 37713&VW). Accessed 24 July 2009. Cliometrica 2007;1:211–37. 32. Buuren van S, Groothuis-Oudshoorn K. MICE: multivariate imputation by 12. Berkey CS, Colditz GA, Rockett HR, Frazier AL, Willett WC. Dairy con- chained equations in R. J Stat Softw 2011;45:1–67. sumption and female height growth: prospective cohort study. Cancer Epi- 33. Buuren van S. Selection issues. In: Buuren van S, ed. Flexible Imputa- demiol Biomarkers Prev 2009;18:1881–7. tion of Missing Data. Boca Raton, FL: Chapman & Hall/CRC Press, 13. Productschap Zuivel (Dutch Dairy Board). EU and other countries - con- 2012:212–8. sumption. In: Productschap Zuivel (Dutch Dairy Board), ed. [Statisti- 34. Sociaal en Cultureel Planbureau (Netherlands Institute for Social Reseach). cal Yearbook 2010]. Zoetermeer, The Netherlands: Productschap Zuivel [Status scores 2006]. (http://www.scp.nl/Organisatie/Onderzoeksgroepen/ (Dutch Dairy Board), 2011:83–107. Wonen_Leefbaarheid_Veiligheid/Lopend_onderzoek_van_WLV/Statuss- 14. Schönbeck Y, Talma H, van Dommelen P, et al. Increase in preva- cores). Accessed 11 January 2010. lence of overweight in Dutch children and adolescents: a compari- 35. Cole TJ, Green PJ. Smoothing reference centile curves: the LMS method son of nationwide growth studies in 1980, 1997 and 2009. PLoS ONE and penalized likelihood. Stat Med 1992;11:1305–19. 2011;6(11):e27608. 36. Buuren van S, Fredriks M. Worm plot: a simple diagnostic device for mod- 15. Freedman DS, Khan LK, Serdula MK, Dietz WH, Srinivasan SR, Berenson elling growth reference curves. Stat Med 2001;20:1259–77. GS. Relation of age at menarche to race, time period, and anthropometric 37. Stasinopoulos DM, Rigby RA. Generalized additive models for location dimensions: the Bogalusa Heart Study. Pediatrics 2002;110:e43. scale and shape (GAMLSS) in R. J Stat Softw 2007;23:1–46.